Alzheimer's disease (AD) is heritable but known risk genes (e.g. APOE 54) explain less than 50% of the observed genetic variance. One reason may be the late onset of symptoms in genetically susceptible persons. A novel solution would be to assess the common genetic variation underlying both incident AD in older adults, and quantitative, heritable, endophenotypes in middle-aged adults. Endophenotypes that demonstrate a graded association with AD risk include MRI total brain and hippocampal volumes [TBV, HV] and performance on verbal memory tests (VM). Other causes of failure to detect genes with moderate effects include small samples, incorrect selection of candidate genes, and the impact of environmental factors. In recent years high-density genome-wide association studies (GWAS) have permitted a successful agnostic approach to the identification of common genetic variants underlying complex diseases such as diabetes and coronary heart disease. The Framingham Heart Study (FHS) has evaluated over 5000 participants for incident dementia and AD. Cognitive tests and brain MRI have been obtained since 1975 and 1999, respectively. Four other large, prospective, epidemiological studies, the Rotterdam (RS), Cardiovascular Health (CHS), Atherosclerosis Risk in Communities (ARIC) and Age, Gene/Environment Susceptibility (AGES) studies have data on incident dementia &AD and/or AD related endophenotypes. All 5 studies will have independently funded GWAS in 2008. Hence we have formed a multinational collaboration and propose a meta-analysis of data on 38,500 participants from these 5 discovery cohorts. This will be followed by Stage 2 replication in 5 external samples (a cost-effective approach to further improve power) and a study of gene-gene and gene-environment interactions. The replication samples will consist of the MIRAGE (Multi-Institutional Research in Alzheimer Genetic Epidemiology), ULSAM (Uppsala Longitudinal Study of Adult Men) and TASCOG (Tasmanian Cognition and Gait) studies and the TGen and Glaxo-Smith-Kline public databases. We propose the following specific aims: Aim 1: To uncover common genetic variation in the discovery cohorts underlying incident dementia (1553 cases, 17,656 at risk), incident AD (1219 cases), and MRI (n=10,711) and VM (n=31,072) endophenotypes. Aim 2: To replicate the 250 strongest SNP-trait associations in the 5 replication samples (2714 AD cases, 3593 controls, 1200 persons with endophenotypes) using in-silico comparisons where feasible and targeted genotyping when required. Aim 3: To examine gene-gene and gene environment interactions modifying the association of 15 selected SNPs with AD and endophenotypes. We will specifically look for epistatic interactions with APOE 54 and GEI interactions with modifiable risk factors such as diet, physical activity, blood pressure and diabetes. We submit that the identification of novel AD genes will provide valuable insights into pathophysiology, and may identify molecular targets for risk stratification and the development of therapies. PUBLIC HEALTH RELEVANCE: We propose a multinational, collaborative effort that will combine genetic information on over 38,500 persons with careful neurological follow-up, MRI and cognitive tests to identify genes causing Alzheimer's disease, the leading cause of dementia. We will identify probable culprit genes in part of the sample, confirm our findings in the remaining persons, and will then examine how these genes interact with each other and with lifestyle factors. We believe our analyses will identify new genes and pathways underlying the risk of Alzheimer's disease, and this in turn may improve our understanding of the disease guiding us to new preventive strategies and treatments.